How ranjeetchauhan0805/retail-sales-and-demand-forecaster works
This is a retail transaction quantity-prediction project: users enter transaction, location, and date details in the Streamlit app and receive a predicted quantity with a summary. The README documents an offline workflow from transaction data through preprocessing, feature engineering, model comparison, and saved-model deployment. Notebook implementation details were not sampled, so those workflow links reflect README documentation; the app’s input, model loading, prediction, and display flow is supported by app.py.
Main entry points: app.py Quantity inference · EDA.ipynb Exploration · gradient_boosting_model.pkl Gradient Boosting model · Sales Dataset.csv Sales transactions
Components 11
Prediction app
- Transaction entry
app.py - Quantity inference
app.py - Prediction and metrics
app.py
Model development
- Exploration
notebook/EDA.ipynb - Preparation and features
notebook/EDA.ipynb - Regression training
notebook/EDA.ipynb - Chronological validation
notebook/EDA.ipynb - Model evaluation
notebook/EDA.ipynb
Data and artifacts
- Sales transactions
data/Sales Dataset.csv - Gradient Boosting model
models/gradient_boosting_model.pkl
Other components
- User
Connections 10
- User → Transaction entry: enters details
- Transaction entry → Quantity inference: submits features
- Quantity inference → Gradient Boosting model: loads and predicts
- Quantity inference → Prediction and metrics: displays prediction
- Sales transactions → Exploration: analyzed in
- Exploration → Preparation and features: informs
- Preparation and features → Regression training: prepares features
- Regression training → Chronological validation: evaluated with
- Chronological validation → Model evaluation: compared in
- Model evaluation → Gradient Boosting model: saves selected model